November 6, 20178 yr So I'm upgrading my cache SSD from SATA to NVMe (an Intel 750). I understood the wiki has a set of instruction involving moving between cache and array (so the old cache will be cleared). However, I would like to still keep the data on the old SATA SSD cache, in case I screw up big time. So can I do this? Plugin the new SSD Format and mount the new SSD as BTRFS using Unassigned Devices Rsync /mnt/cache/* to /disks/NVMe Start new config, keeping array and parity Assign NVMe to the cache slot I reckon the above would be faster due to (1) only 1 copying between SSD instead of SSD -> HDD -> SSD and (2) not having to involve the slow array. The only slight concern is rsync and how it deals with links. Hopefully it doesn't cause my appdata to blow up in size.
November 6, 20178 yr Community Expert It works as long as VM and docker are stopped as well, you don't need to do a new config, when the copy is done just stop the array and assign the new cache device. If your current SSD is btrfs you can also do an online replacement, i.e., you don't even have to stop your docker/VMs, but data will be moved from the old to the new device, so it's not what you want.
November 7, 20178 yr Author 9 hours ago, johnnie.black said: It works as long as VM and docker are stopped as well, you don't need to do a new config, when the copy is done just stop the array and assign the new cache device. If your current SSD is btrfs you can also do an online replacement, i.e., you don't even have to stop your docker/VMs, but data will be moved from the old to the new device, so it's not what you want. Thanks Johnnie. Just out of curiosity, is the "online replacement" = add disk to cache pool, balance, then remove old disk? I remember there was a command that need to be typed into the console before removing the old disk which cleared the data. Wondering what would happen if I only balance the pool and remove the old disk without running that command (or run the command after removing the disk). Theoretically, a 2-disk cache pool is a RAID 1 which is mirror so removing 1 disk should not cause any problem with data.
November 7, 20178 yr Community Expert 1 minute ago, testdasi said: Just out of curiosity, is the "online replacement" = add disk to cache pool, balance, then remove old disk? No, it's using the replace command:
November 9, 20178 yr Author Just want to add 1 thing in case anyone found this useful. I had to recreate my Docker image due to the NOCOW bit thingie (the warning looks serious but a bit google says it's only a problem in a certain scenario - I recreate nevertheless). I used the below command to copy data so that may be the cause but no big deal. rsync -avXHq /mnt/cache/docker /mnt/disks/Intel750 Otherwise it has been pretty uneventful so far. *fingers crossed* Edited November 9, 20178 yr by testdasi
Archived
This topic is now archived and is closed to further replies.